Back to the stories

AMD reports record data-center revenue and scales Helios rack-scale AI systems

Score 10.0,

AI Infrastructure and Compute

AI infrastructure competition is moving from individual chips to complete systems already carrying real workloads.

AMD’s data-center business hit a record, with revenue up 97% year over year. The company says major technology firms are running workloads on its first Helios systems, including Meta and OpenAI.

Why does that matter? Building more capable AI takes more than clever models. It also takes a growing supply of powerful computers. AMD is showing it can deliver a complete system, not just individual chips.

Helios brings AI-focused chips and server processors together in a rack, designed to work as one large computing system. Think of it as a workshop full of machines wired to work together, rather than one computer doing the whole job alone.

For companies building AI services, that gives them another complete system to test and potentially deploy. But these are early systems. AMD says production will scale further in the fourth quarter, so initial workloads are not yet proof of broad availability or better performance.

The shift is striking: competition is now being tested in data centers, with real workloads. The next question is whether AMD can turn those first deployments into a steady supply of systems that customers choose at scale.